projects/experimental/ner_spancat_compare/project.yml at v3 · explosion/projects · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
title: "Comparing SpanCat and NER using a corpus of biomedical literature (GENIA)"
description: |
  This project demonstrates how spaCy's Span Categorization (SpanCat) and
  Named-Entity Recognition (NER) perform on different types of entities. Here, we used
  a dataset of biomedical literature containing both overlapping and non-overlapping spans.

vars:
  name: "ner_spancat_compare"
  version: "1.0.0"
  gpu_id: -1
  spans_key: "sc"
  # Labels
  dna: "DNA"
  rna: "RNA"
  cell_line: "cell_line"
  cell_type: "cell_type"
  protein: "protein"

directories:
  - "assets"
  - "configs"
  - "corpus"
  - "corpus/ner"
  - "metrics"
  - "scripts"
  - "training"

workflows:
  all:
    - "download"
    - "convert"
    - "create-ner"
    - "train-ner"
    - "assemble-ner"
    - "train-spancat"
    - "evaluate-ner"
    - "evaluate-spancat"
  spancat:
    - "download"
    - "convert"
    - "train-spancat"
    - "evaluate-spancat"
  ner:
    - "download"
    - "convert"
    - "create-ner"
    - "train-ner"
    - "evaluate-ner"
    - "assemble-ner"

assets:
  - dest: "assets/train.iob2"
    description: "The training dataset for GENIA in IOB format."
    url: https://github.com/thecharm/boundary-aware-nested-ner/blob/master/Our_boundary-aware_model/data/genia/genia.train.iob2
  - dest: "assets/dev.iob2"
    description: "The evaluation dataset for GENIA in IOB format."
    url: https://github.com/thecharm/boundary-aware-nested-ner/blob/master/Our_boundary-aware_model/data/genia/genia.dev.iob2
  - dest: "assets/test.iob2"
    description: "The test dataset for GENIA in IOB format."
    url: https://github.com/thecharm/boundary-aware-nested-ner/blob/master/Our_boundary-aware_model/data/genia/genia.test.iob2

commands:
  - name: "download"
    help: "Download model-related assets"
    script:
      - "python -m spacy download en_core_web_lg"
  - name: "convert"
    help: "Convert IOB file into the spaCy format"
    script:
      - "python -m scripts.convert assets/train.iob2 -o corpus/train.spacy"
      - "python -m scripts.convert assets/dev.iob2 -o corpus/dev.spacy"
      - "python -m scripts.convert assets/test.iob2 -o corpus/test.spacy"
    deps:
      - "assets/train.iob2"
      - "assets/dev.iob2"
      - "assets/test.iob2"
    outputs:
      - "corpus/train.spacy"
      - "corpus/dev.spacy"
      - "corpus/test.spacy"

  - name: "create-ner"
    help: "Split corpus into separate NER datasets for each GENIA label"
    script:
      - >-
        python -m scripts.create_ner
        --train corpus/train.spacy
        --dev corpus/dev.spacy
        --test corpus/test.spacy
        --output-dir corpus/ner
    deps:
      - "corpus/train.spacy"
      - "corpus/dev.spacy"
      - "corpus/test.spacy"
    outputs:
      - "corpus/ner/train_${vars.dna}.spacy"
      - "corpus/ner/train_${vars.rna}.spacy"
      - "corpus/ner/train_${vars.cell_line}.spacy"
      - "corpus/ner/train_${vars.cell_type}.spacy"
      - "corpus/ner/train_${vars.protein}.spacy"
      - "corpus/ner/dev_${vars.dna}.spacy"
      - "corpus/ner/dev_${vars.rna}.spacy"
      - "corpus/ner/dev_${vars.cell_line}.spacy"
      - "corpus/ner/dev_${vars.cell_type}.spacy"
      - "corpus/ner/dev_${vars.protein}.spacy"
      - "corpus/ner/test_${vars.dna}.spacy"
      - "corpus/ner/test_${vars.rna}.spacy"
      - "corpus/ner/test_${vars.cell_line}.spacy"
      - "corpus/ner/test_${vars.cell_type}.spacy"
      - "corpus/ner/test_${vars.protein}.spacy"

  - name: "train-ner"
    help: "Train an NER model for each label"
    script:
      - >-
        python -m spacy train
        configs/ner.cfg
        --output training/ner/${vars.dna}
        --paths.train corpus/ner/train_${vars.dna}.spacy
        --paths.dev corpus/ner/dev_${vars.dna}.spacy
        --gpu-id ${vars.gpu_id}
      - >-
        python -m spacy train
        configs/ner.cfg
        --output training/ner/${vars.rna}
        --paths.train corpus/ner/train_${vars.rna}.spacy
        --paths.dev corpus/ner/dev_${vars.rna}.spacy
        --gpu-id ${vars.gpu_id}
      - >-
        python -m spacy train
        configs/ner.cfg
        --output training/ner/${vars.cell_line}
        --paths.train corpus/ner/train_${vars.cell_line}.spacy
        --paths.dev corpus/ner/dev_${vars.cell_line}.spacy
        --gpu-id ${vars.gpu_id}
      - >-
        python -m spacy train
        configs/ner.cfg
        --output training/ner/${vars.cell_type}
        --paths.train corpus/ner/train_${vars.cell_type}.spacy
        --paths.dev corpus/ner/dev_${vars.cell_type}.spacy
        --gpu-id ${vars.gpu_id}
      - >-
        python -m spacy train
        configs/ner.cfg
        --output training/ner/${vars.protein}
        --paths.train corpus/ner/train_${vars.protein}.spacy
        --paths.dev corpus/ner/dev_${vars.protein}.spacy
        --gpu-id ${vars.gpu_id}
    deps:
      - "corpus/ner/train_${vars.dna}.spacy"
      - "corpus/ner/train_${vars.rna}.spacy"
      - "corpus/ner/train_${vars.cell_line}.spacy"
      - "corpus/ner/train_${vars.cell_type}.spacy"
      - "corpus/ner/train_${vars.protein}.spacy"
      - "corpus/ner/dev_${vars.dna}.spacy"
      - "corpus/ner/dev_${vars.rna}.spacy"
      - "corpus/ner/dev_${vars.cell_line}.spacy"
      - "corpus/ner/dev_${vars.cell_type}.spacy"
      - "corpus/ner/dev_${vars.protein}.spacy"
    outputs:
      - "training/ner/${vars.dna}/model-best"
      - "training/ner/${vars.rna}/model-best"
      - "training/ner/${vars.cell_line}/model-best"
      - "training/ner/${vars.cell_type}/model-best"
      - "training/ner/${vars.protein}/model-best"

  - name: "train-spancat"
    help: "Train a SpanCat model"
    script:
      - >-
        python -m spacy train
        configs/spancat.cfg
        --output training/spancat/
        --paths.train corpus/train.spacy
        --paths.dev corpus/dev.spacy
        --gpu-id ${vars.gpu_id}
    deps:
      - "corpus/train.spacy"
      - "corpus/dev.spacy"
      - "corpus/test.spacy"
    outputs:
      - "training/spancat/model-best"

  - name: "evaluate-ner"
    help: "Evaluate all NER models"
    script:
      - "mkdir -p metrics/ner"
      - >-
        python -m spacy evaluate
        training/ner/${vars.dna}/model-best
        corpus/ner/test_${vars.dna}.spacy
        --output metrics/ner/scores_${vars.dna}.json
      - >-
        python -m spacy evaluate
        training/ner/${vars.rna}/model-best
        corpus/ner/test_${vars.rna}.spacy
        --output metrics/ner/scores_${vars.rna}.json
      - >-
        python -m spacy evaluate
        training/ner/${vars.cell_line}/model-best
        corpus/ner/test_${vars.cell_line}.spacy
        --output metrics/ner/scores_${vars.cell_line}.json
      - >-
        python -m spacy evaluate
        training/ner/${vars.cell_type}/model-best
        corpus/ner/test_${vars.cell_type}.spacy
        --output metrics/ner/scores_${vars.cell_type}.json
      - >-
        python -m spacy evaluate
        training/ner/${vars.protein}/model-best
        corpus/ner/test_${vars.protein}.spacy
        --output metrics/ner/scores_${vars.protein}.json
    deps:
      - "training/ner/${vars.dna}/model-best"
      - "training/ner/${vars.rna}/model-best"
      - "training/ner/${vars.cell_line}/model-best"
      - "training/ner/${vars.cell_type}/model-best"
      - "training/ner/${vars.protein}/model-best"
      - "corpus/ner/test_${vars.dna}.spacy"
      - "corpus/ner/test_${vars.rna}.spacy"
      - "corpus/ner/test_${vars.cell_line}.spacy"
      - "corpus/ner/test_${vars.cell_type}.spacy"
      - "corpus/ner/test_${vars.protein}.spacy"
    outputs:
      - "metrics/ner/scores_${vars.dna}.json"
      - "metrics/ner/scores_${vars.rna}.json"
      - "metrics/ner/scores_${vars.cell_line}.json"
      - "metrics/ner/scores_${vars.cell_type}.json"
      - "metrics/ner/scores_${vars.protein}.json"

  - name: "assemble-ner"
    help: "Assemble all NER models into a single pipeline"
    script:
      - >-
        python -m spacy assemble
        configs/ner_assemble.cfg
        training/ner-assemble/
        --code scripts/transfer_ent_component.py
        --paths.ner_${vars.dna} training/ner/${vars.dna}/model-best
        --paths.ner_${vars.rna} training/ner/${vars.rna}/model-best
        --paths.ner_${vars.cell_line} training/ner/${vars.cell_line}/model-best
        --paths.ner_${vars.cell_type} training/ner/${vars.cell_type}/model-best
        --paths.ner_${vars.protein} training/ner/${vars.protein}/model-best
        --paths.spans_key ${vars.spans_key}
        --verbose
    outputs:
      - "training/ner-assemble/"

  - name: "evaluate-spancat"
    help: "Evaluate SpanCat model"
    script:
      - "mkdir -p metrics/spancat"
      - >-
        python -m spacy evaluate
        training/spancat/model-best
        corpus/test.spacy
        --output metrics/spancat/scores.json
    deps:
      - "training/spancat/model-best"
      - "corpus/test.spacy"
    outputs:
      - "metrics/spancat/scores.json"